KR20070001297A - Method of embedding and extracting watermark using wavelet transform - Google Patents

Abstract

A method for inserting and detecting a watermark by using a wavelet transformation is provided to satisfy the invisibility and robustness. An insertion image is converted by a discrete wavelet conversion so that plural sub bands are formed. The information on each insertion area is decoded through predetermined encryption algorithm. The watermark is extracted from the insertion area on the basis of the decoded information on the insertion area. At the watermark insertion method, the wavelet conversion by means of lifting is used. The lifting conversion is implemented at the spatial shaft of bi-orthogonal wavelet conversion. The wavelet conversion using the lifting includes split, predict and update processes.

Description

Watermark embedding and detection using wavelet transform {METHOD OF EMBEDDING AND EXTRACTING WATERMARK USING WAVELET TRANSFORM}

1 is a control block diagram of a lifting converter and a reverse lifting converter according to the present invention,

2 is a diagram showing the structure of a subband formed by wavelet transform according to the present invention,

3 is a diagram illustrating a watermark embedding algorithm according to the present invention;

4 is a diagram illustrating a watermark extraction algorithm according to the present invention.

* Explanation of symbols for the main parts of the drawings

10a: lifting conversion unit 10b: reverse lifting conversion unit

The present invention relates to a watermark embedding and detection method using wavelet transform, and more particularly, to insert a watermark in a wavelet transform region having spatial domain information and frequency domain information, and detect a watermark without an original image. The present invention relates to a watermark embedding and detection method using wavelet transform.

With the development of the network and the Internet, the commercial model of digitized content distribution has been established, and the necessity of copyright protection technology of digital content is also emerging.

To solve this problem, digital image watermarking technology that inserts specific data that is intended to claim copyright information, such as copyright information, into multimedia contents cannot be distinguished by the human eye or hearing, and is recognized as a copyright protection solution. It is recognized as an essential technology for establishing standardized copyright information and distributing healthy digital contents.

In particular, the digital image containing the most implicit information is a complicated process and requires a large amount of computation, so the wittmarking technique that satisfies both the invisibility and the robustness of the watermark through image processing at the same time Needs a lot of research.

In the digital image domain, watermarking, which is mainly used in the spatial domain, is being moved to the research domain due to its application in the frequency domain. The watermarking is more resistant to attack than the method applied in the spatial domain. It has a disadvantage that cannot be selected.

However, the introduction of the wavelet region, which has both the characteristics of the frequency domain and the spatial domain, makes the insertion of watermarks more efficient.

Watermarking in the frequency domain of a digital image inserts a watermark by changing a frequency coefficient. In this regard, "Secure Spread Spectrum Watermarking for Multimedia" (IJ Cox, J. Killian, T. Leighton and T. Shamoon, IEEE Trans. On Image Processing, 6, 12, pp. 1673-1687, 1997) and "A Review of Watermarking and the Importance of Perceptual Modeling "(IJ Cox and ML Miller, Proc. Of SPIE Conf. On Human Vision and Electronic Imaging II, vol. 3016, pp. 92-99, February, 1997). A method of extracting an important coefficient in the frequency domain using a transform and inserting a watermark is disclosed.

In addition, in the "Image Watermarking of Secure Transmission over Public Networks" (M. Barni, Proc. Of COST 254 Workshop on Emerging Techniques for Communication Terminals, Toulouse, France, pp. 290-294, July, 1997) ), A method of embedding a watermark in a phase is disclosed, and "A Multiresolution Watermark for Digital Images" (XG Xia, CG Boncelet and GR Arce, Proc. Of IEEE ICIP, vol. 3, pp. 548-551, 1997.) and "A Multiresolution Watermark for Digital Images" (XG Xia, CG Boncelet and GR Arce, Proc. Of IEEE ICIP, vol. 3, pp. 548-551, 1997.). A watermark method using multiresolution is disclosed.

In such a conventional watermark embedding and extraction method, an original image is required when a watermark is detected in a digital image into which a watermark has been inserted.

Accordingly, an object of the present invention is to insert a watermark in a wavelet transform region having spatial domain information and frequency domain information, and to detect the watermark without the original image, while satisfying invisibility and robustness. The present invention provides a method of embedding and extracting watermarks of digital images using wavelet transform.

According to the present invention, there is provided a watermark embedding method using wavelet transform, comprising: forming a plurality of subbands by performing discrete wavelet transform of an original image; Selecting at least one subband from which the watermark is to be inserted among the plurality of subband areas; Calculating a wavelet coefficient for the selected subband based on a predetermined threshold value, and selecting an insertion region into which a watermark is to be inserted on the selected subband based on the calculated correlation between the wavelet coefficient and the subbands Steps; Encrypting information on the insertion region through a predetermined encryption algorithm; And embedding the watermark based on the wavelet coefficients into predetermined adjustment values to generate an embedded image in which the watermark is embedded.

The selecting of the insertion region may include extracting a first wavelet coefficient greater than the threshold value for the selected subband; Calculating a predetermined first threshold value based on a deviation between an average of four first wavelet coefficients adjacent to any one of the selected subbands and a wavelet coefficient of the subbands corresponding to the four first wavelet coefficients; ; The method may include selecting the insertion region based on the threshold value, the first threshold value, and the correlation.

The forming of the plurality of subbands may be performed by a lifting transform algorithm.

The predetermined encryption algorithm may include a block encryption algorithm (SEED).

On the other hand, the object is a watermark extraction method for extracting a watermark from the embedded image generated by the watermark embedding method according to another embodiment of the present invention, a plurality of units by discrete wavelet transforming the embedded image Forming an inverse; Decrypting the information about the insertion region through the predetermined encryption algorithm; And extracting the watermark from the embedding area based on the information on the complexed embedding area.

Hereinafter, with reference to the accompanying drawings will be described in detail with respect to the present invention.

First, the watermark embedding method according to the present invention uses wavelet transform using lifting. 1 is a control block diagram of a lifting conversion unit 10a according to the present invention.

Referring to FIG. 1, the lifting transform is implemented on the spatial axis of the Bi-orthogonal wavelet transform. As shown in FIG. 1, wavelet transform using lifting includes splitting 11a (S: Split), prediction 12a (P: Predict), and updating 13a (U: Update).

The dividing step divides the input signal x [n] into two components, an even signal x _e [n] and an odd signal x _o [n].

In the prediction step, the prediction operator P obtains an error wavelet coefficient d [n] generated when x _o [n] is predicted from x _e [n]. Here, the error wavelet coefficient d [n] is calculated through Equation 1.

[Equation 1]

d [n] = x _o [n]-P (x _e [n])

The update step combines x _e [n] and d [n] to obtain a scaling factor c [n] that approximates the input signal x [n]. The scaling factor c [n] is calculated through Equation 2.

[Equation 2]

c [n] = x _e [n]-U (d [n])

In the lifting transformation according to the present invention, a Daubechies filter may be used. Here, the factoring disclosed in "Factoring wavelet and subband transforms into lifting steps" (I. Daubechies and W. Sweldens, Journal of Fourier Analysis and Applications , 4 (3): 245-267, 1998.) If we convert to lifting by applying (Factoring) algorithm, we can get the following result when input signal is x _i .

s _i ⁽⁰⁾ = x _2i ,

d _i ⁽⁰⁾ = x _{2i + 1} ,

d _i ⁽¹⁾ = d _i ⁽⁰⁾ + α (s _i ⁽⁰⁾ + s _{i + 1} ⁽⁰⁾ ),

s _i ⁽¹⁾ = s _i ⁽⁰⁾ + β (d _i ⁽¹⁾ + d _i-1 ⁽¹⁾ ),

d _i ⁽²⁾ = d _i ⁽¹⁾ + γ (s _i ⁽¹⁾ + s _{i + 1} ⁽¹⁾ ),

s _i ⁽²⁾ = s _i ⁽¹⁾ + δ (d _i ⁽²⁾ + d _i-1 ⁽²⁾ ),

s _i = ζs _i ⁽²⁾ ,

d _i = d _i ⁽²⁾ ζ

Here, α is -1.586134342, β is -0.052980118, γ is 0.8829110762, δ is 0.4435068522, ζ is arbitrarily defined as being 1.149604398.

Then, the input signal x _i is divided by the even-numbered signal s _i ⁽⁰⁾ and the odd-numbered signal d _i ^(0), and then the predicted value d using the neighboring even-numbered signals s _i ⁽⁰⁾ and s _{i + 1} ⁽⁰⁾ . _{Find i} ⁽¹⁾ .

In the next update step, s _i ⁽¹⁾ is found using two neighboring odd _- numbered signals d _i-1 ⁽¹⁾ and d _i ⁽¹⁾ . The prediction and update process is performed one step further, and the resultant values s _i ⁽²⁾ and d _i ⁽²⁾ are multiplied by the scaling factor ζ, respectively, to obtain final coefficient values s _i and d _i .

Unexplained reference numeral 10b in FIG. 1 is an inverse lifting transform section 11b, which corresponds to the configuration of the lifting transform section 11a, and updates 13b, U, prediction 12a, P, and merge 11b. It consists of (M: Merge).

The wavelet transform is performed through the above process. In the present invention, the three-level lifting transform is performed as an example. Accordingly, a subband structure as shown in FIG. 2 is obtained.

As shown in FIG. 2, subbands of LL, LH, HL, and HH are generated as a result of the conversion for each step, and each subband includes a specific frequency component of the original image.

Here, LL denotes a low frequency component, LH denotes a high frequency component having a horizontal frequency characteristic of an image, HL denotes a high frequency component having a vertical frequency characteristic, and HH denotes a high frequency component having a diagonal frequency characteristic.

Here, since down-sampling is performed during the lifting transform, one wavelet coefficient at any position in step j correlates with four coefficients in step i-1. The watermark embedding method according to the present invention performs watermark embedding and detection using the correlation between the subbands as described above.

In the present invention, as an example, the selection of the LH region is selected as a position for inserting a watermark from each subband generated by wavelet transform. Here, the selection of the LH region may be selected according to the result of examining the robustness of each subband by performing JPEG compression of several steps after lifting conversion on 500 still images. That is, it examines the robustness of each subband for general image processing such as JPEG, and an example thereof is shown in Table 1 below.

TABLE 1

JPEG Quality Error Ratio (%) LL3 LH3 HL3 HH3 12 0.0 0.0 0.0 0.0 10 0.0 0.0 0.0 0.5 8 0.0 0.3 0.3 20.9 6 0.0 3.6 3.3 34.3 4 2.9 23.8 24.7 55.5 2 1.6 43.0 48.8 69.2 0 36.7 58.2 65.4 76.1

As shown in Table 1, the subbands of LH3 show the most robust characteristics except LL3. In particular, it can be seen that the subband of LH3 is more robust as the compression ratio increases.

Meanwhile, the watermark embedding method according to the present invention inserts the watermark only in a predetermined region on the selected subband through the above process. Hereinafter, a process of selecting a watermark insertion region in the watermark embedding method according to the present invention will be described by defining an area on the selected subband in which the watermark is inserted as an insertion region.

In general, an encoding scheme is used for efficient transmission of video over a network. Among these coding techniques, "Embedded image coding using zerotrees of wavelet coefficients" (Shapiro, JM, Signal Processing, IEEE Transactions on, Volume: 41 Issue: 12, Dec. 1993, pp 3445 -3462) is used for progressive transmission of images. "A New, Fast and Efficient Image codec Based on Set Partitioning in Hierarchical Trees" (Amir Said and William A. Pearlman, IEEE Trans. On Circuits and Systems for Video Technology, Vol. 6. No. 3, pp. 243-250, June 1996.) discloses a method of improving a data transmission rate with a high compression rate and good image quality.

The above-described algorithms generate an embedded bitstream by first sending significant bits by partial alignment, and compress the image by zero-tree coding using similarity between subbands in the wavelet transformed image.

The method of selecting the critical coefficients in the zero tree coding to select the critical coefficients of the intermediate frequency band predicted from the high frequency subbands in the wavelet region and insert the watermark is "Hiding digital watermarks using multiresolution wavelet transform" (Ming -Shing Hsieh, Din-Chang Tseng, Yong-Huai Huang, Industrial Electronics, IEEE Transactions on, Volume: 48 Issue: 5, Oct. 2001, pp 875 -882. The watermark embedding method according to the present invention improves this to determine the watermark embedding position and to insert the watermark.

First, the wavelet coefficients larger than the respective thresholds T ₁ and T ₂ are extracted from the selected subbands LH1 and LH2. Here, the threshold value is used by applying a predetermined amount of scaling to a median value in the selected subband.

Here, by giving an arbitrary weight to the median, the median is chosen from a value of absolute values 0 to 5 to prevent the deterioration of the watermarking efficiency caused by too many coefficients or too small coefficients being selected.

The wavelet coefficient correlation between LH2 and LH3 is then examined for the selected coefficients. The correlation is then calculated as the difference between the average of four adjacent coefficient values of LH2 and the coefficient values of LH3 corresponding to these coefficients, and the generated difference values are again selected by the threshold value T ₃ .

Since the positions selected by the thresholds T ₁ and T ₂ exceed the number of watermarks to be inserted (32 × 32) corresponding to approximately 2000, an experimental threshold T ₃ is introduced for the adjustment thereof.

The coefficients selected by the three thresholds correspond to the positions where the watermark is to be inserted, and these positions are encrypted with a block cipher algorithm (SEED) for security. Here, the three thresholds become keys for watermark extraction later.

Then, the insertion position P for inserting the watermark is selected using the correlation between the threshold value and the subband.

[Equation 3]

Hereinafter, a watermark embedding method according to the present invention will be described with reference to FIG. In this case, the original image uses a 512 ㅧ 512 size, and the watermark image uses a 32 ㅧ 32 size binary image. In addition, a scaling component T ₄ for controlling the size of the inserted watermark information is introduced and used as a threshold.

First, the coefficient S _{3, j} of the position selected to insert the watermark is adjusted as follows by ?? and T3.

The adjusted value in the above process is readjusted by T ₄ and diff, and the watermark is inserted by weighting the original value according to the watermark to a positive or negative value.

After the watermark embedding process is completed, the watermark-embedded image I 'is obtained through inverse lifting conversion.

In this case, the information P indicating the position of the insertion region where the watermark is inserted is encrypted and stored by a block cipher algorithm (SEED) as an example. This encrypted information P 'is used for extraction of the watermark.

3 is a diagram illustrating a watermark embedding algorithm according to the present invention through the above process.

Hereinafter, a watermark extraction process according to the present invention will be described. 4 is a diagram illustrating a watermark extraction algorithm according to the present invention.

The watermark extraction method according to the present invention detects the watermark using the position information P of the inserted watermark without using the original image. As described above, the encrypted information P 'is decrypted according to the block cipher algorithm SEED to extract the position where the watermark is inserted.

Here, the watermark extraction process according to the present invention corresponds to an inverse process of the above-described watermark embedding process.

First, three-level lifting conversion is performed on the image I 'with the watermark embedded therein. Then, the encrypted binary permutation (P ′) is decrypted to extract information (P) about the location of the watermark, and the subbands (S _{3, j} ) and the threshold value (T ₄ ) are determined to extract the watermark. Detect information for

Then, calculate the diff and finally extract the watermark (W) by comparing diff with T ₄ .

 Although some embodiments of the invention have been shown and described, it will be apparent to those skilled in the art that modifications may be made to the embodiment without departing from the spirit or spirit of the invention. . And the scope of the invention will be defined by the appended claims and equivalents thereof.

As described above, according to the present invention, the watermark is inserted in the wavelet transform region having the spatial domain information and the frequency domain information, and the invisibility and robustness can be detected while the watermark can be detected without the original image. A watermark embedding and extraction method of digital image using wavelet transform is provided.

Claims (5)

In the watermark embedding method using wavelet transform, Discrete wavelet transforming the original image to form a plurality of subbands; Selecting at least one subband from which the watermark is to be inserted among the plurality of subband areas; Calculating a wavelet coefficient for the selected subband based on a predetermined threshold value, and selecting an insertion region into which a watermark is to be inserted on the selected subband based on the calculated correlation between the wavelet coefficient and the subbands Steps; Encrypting information on the insertion region through a predetermined encryption algorithm; And inserting a watermark based on the wavelet coefficients based on predetermined adjustment values to generate an embedded image in which the watermark is inserted. The method of claim 1, Selecting the insertion region, Extracting a first wavelet coefficient greater than the threshold for the selected subband; Calculating a predetermined first threshold value based on a deviation between an average of four first wavelet coefficients adjacent to any one of the selected subbands and a wavelet coefficient of the subbands corresponding to the four first wavelet coefficients; ; And selecting the insertion area based on the threshold value, the first threshold value, and the correlation. The method of claim 2, And forming the plurality of subbands is performed by a lifting conversion algorithm. The method of claim 3, And said predetermined encryption algorithm comprises a block encryption algorithm (SEED). A watermark extraction method for extracting a watermark from an embedded image generated by the watermark embedding method according to claim 4, Discrete wavelet transforming the inserted image to form a plurality of subbands; Decrypting the information about the insertion region through the predetermined encryption algorithm; And extracting the watermark from the embedding area based on the information on the complexed embedding area.

Images